Sijia Liu

Excavation of Molecular Subtypes of Cervical Cancer Based on DNA Methylation Patterns

Background: Cervical cancer remains a major cause of cancer-related death among women worldwide. Despite advances in treatment, prognosis remains poor for many patients due to tumor heterogeneity. DNA methylation, an epigenetic modification, is known to influence tumor development, but its role in defining molecular subtypes and prognostic stratification in cervical cancer remains inadequately understood. Methods: We analyzed DNA methylation profiles from 287 cervical cancer samples obtained from the UCSC Xena database. Univariate and multivariate Cox regression analyses were applied to identify prognostic CpG sites, as these models allow evaluation of individual and combined effects of methylation sites on patient survival. Consensus clustering was performed to define robust molecular subtypes based on methylation patterns, providing insights into tumor heterogeneity. Differentially methylated regions were identified using the Quantitative Differentially Methylated Regions (QDMR) software, an entropy-based tool validated for detecting subtype-specific methylation markers. A Bayesian classifier was constructed and validated in training and test cohorts to evaluate the predictive accuracy of these markers for subtype classification. Additionally, immune cell infiltration was estimated using computational algorithms to assess tumor microenvironment differences, and chemosensitivity was predicted to explore potential clinical implications of the methylation subtypes. Results: Four distinct methylation-based subtypes differed in methylation patterns, histological types, clinical stages, and metastatic status. A total of 501 subtype-specific methylation sites were identified. The Bayesian classifier demonstrated strong predictive performance, with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.824 based on 10-fold cross-validation, indicating high classification accuracy and robustness. The immune microenvironment composition varied markedly among subtypes. Notably, Cluster 1 had elevated infiltration of central memory CD8+ and effector memory CD4+ T cells, whereas Cluster 4 exhibited reduced immune activation and the lowest immune checkpoint expression. These findings indicate subtype-specific differences in potential responsiveness to immunotherapy. Conclusions: These DNA methylation-driven subtypes highlight the heterogeneity of cervical cancer and offer new insights for personalized therapy.

Prognostic prediction in primary cervical squamous cell carcinoma with serum squamous cell carcinoma antigen ≥ 10 ng/mL: development and validation of a nomogram model based on inflammatory biomarkers and clinical factors

Cervical cancer remains a significant health burden worldwide, particularly in patients with markedly elevated pretreatment serum squamous cell carcinoma antigen levels (≥ 10 ng/mL), who often have poor outcomes. Accurate prognostic tools for this high-risk population are limited. We conducted a single-center retrospective study including 355 patients with primary cervical squamous cell carcinoma who received radiotherapy between 2020 and 2023. Clinical characteristics and inflammation-related hematological indices, including hemoglobin-to-red cell distribution width ratio and neutrophil-to-lymphocyte ratio, were collected. Univariate and multivariate Cox regression analyses were performed to identify independent prognostic factors for overall survival. A nomogram incorporating these variables, along with clinical stage and treatment modality, was developed and validated. Hemoglobin-to-red cell distribution width ratio, neutrophil-to-lymphocyte ratio, clinical stage, and treatment modality were independent predictors of survival. The nomogram achieved a concordance index of 0.729 in the training cohort and 0.704 in the validation cohort. The area under the time-dependent receiver operating characteristic curves for overall survival at 1, 2, and 3 years were 0.76, 0.81, and 0.79 in the training cohort, and 0.80, 0.75, and 0.71 in the validation cohort. Decision curve analysis demonstrated a consistent net clinical benefit, and risk stratification based on total scores effectively distinguished high- and low-risk groups. This study developed and validated a clinically useful nomogram integrating inflammation-related hematological indices and clinical parameters to predict survival in high-risk cervical cancer patients. The model demonstrates favorable predictive accuracy and may guide individualized treatment strategies, supporting its potential for clinical application.